Keynote Speaker

Keynote Title: What We Learned Measuring Software Energy (and What Still Doesn’t Work)

Abstract: As software scales across clouds, AI pipelines, and embedded systems, its energy consumption has become one of the field's most important and hardest to assess properties. We build systems that consume enormous energy, yet we rarely know how much, why, or what we could do about it.

This talk traces GreenLab's decade-long journey in software energy measurement: from an educational tool to a systematic benchmarking platform spanning code to architecture, culminating in its evolution toward a holistic view of sustainable software. Along the way, we defined tools and protocols to measure it at different granularities, automated large-scale experiments, and adopted hardware probes in contexts where software-level instrumentation falls short.

We reflect on the open challenges: measuring energy in virtualized environments, bridging the gap between academic findings and industrial practice, and modeling software energy behavior. The talk draws on real studies across AI workloads, cloud deployments, and code-level optimizations to show what rigorous energy measurement can and cannot tell us. Finally, we argue about the importance of experimentation on software energy efficiency in academia and the impact on current society.

The audience will leave with a clearer picture of what it takes to measure software energy well, where the field's blind spots still lie, and why closing the gap between research and practice may matter more than any individual measurement result.